When I had looked at this before (a few years back) the ARM compiler beat every other compiler (not surprisingly). And like Marko said, no company I worked for wanted to pay for the compiler even though it did reduce code size between 10 and 15%.
Having a bunch of other compiler support would just add to the testing/support burden. Not sure it is worth it for 10-15% (imo). > On Jul 7, 2017, at 12:35 PM, marko kiiskila <ma...@runtime.io> wrote: > > Hi, > > I suspect that is still true regarding IAR, Keil. YMMV regarding > how much gain you’ll get. > > That being said, I’ve always ended up using gcc. The gains from > commercial compilers have not been worth the seat cost for me. > This given that the size difference in output (as an example) has been > in the ballpark of 5% - 15%. I don’t have any concrete numbers in > front me, it’s been a while since I looked at this (so the ballpark might > or might not be accurate). > >> On Jul 6, 2017, at 11:10 PM, Raoul van Bergen <raoul.van.ber...@icloud.com> >> wrote: >> >> 3 years ago we performed some benchmarking comparing gcc against IAR for a >> cortex M3 design and at that time the IAR outperformed gcc significantly >> both in terms of code speed and size (so optimization). >> If this still holds true is unknown to me and it would be interesting if >> someone could do the comparison on Mynewt with current versions of compilers. >> >> >> Raoul van Bergen >> >>> On 6. Jul 2017, at 22:33, Sterling Hughes >>> <sterling.hughes.pub...@gmail.com> wrote: >>> >>> I think clang support is definitely a requirement going forward. IMO It >>> would be nice to support IAR and Keil as options, although it would mean >>> maintaining multiple linker scripts and a portability layer that abstracted: >>> >>> - Packed structs >>> - Inline assembler >>> - Memory sections >>> >>> Not impossible, but hard to do readably. >>> >>> What are people’s thoughts? I know Keil and IAR have reasonable usage now, >>> but are they important compilers to invest time into now to support, or has >>> their time passed? >>> >>> Sterling >>> >>>> On 6 Jul 2017, at 11:14, Christopher Collins wrote: >>>> >>>>> On Thu, Jul 06, 2017 at 06:41:12PM +0100, Jonathan Pallant wrote: >>>>> Hi, I just wanted to jump in here and suggest that -std=c11 is a better >>>>> choice than -std=gnuXX. If you allow GNU specific extensions then you >>>>> might have issues using other compilers - certainly I would be surprised >>>>> to see the project to rule out clang support in the future. I would >>>>> suggest c11 over c99 as it offers things like atomics which might be >>>>> useful. >>>> >>>> I was under the impression that Mynewt already uses some gnu extensions, >>>> though I could be wrong about that. Regarding clang- it also supports >>>> the gnuXX "standards". It would be nice to support other compilers as >>>> well, but I think that would be quite an uphill battle, and these days, >>>> I wonder how many people would want to use a toolchain other than gcc or >>>> clang for Mynewt. >>>> >>>> Chris >
